See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/260617256

A Review of the Tensile Strength of Rock: Concepts and Testing

Article in Geotechnical and Geological Engineering · January 2014

DOI: 10.1007/s10706-014-9732-0

CITATIONS READS

284 28,280

2 authors:

Matthew A Perras Mark S. Diederichs

York University Queen's University
60 PUBLICATIONS 700 CITATIONS 265 PUBLICATIONS 7,610 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Umbrella Arch View project

The Niagara Tunnel Project View project

All content following this page was uploaded by Matthew A Perras on 21 March 2014.

The user has requested enhancement of the downloaded file.

Geotech Geol Eng (2014) 32:525–546
DOI 10.1007/s10706-014-9732-0
ORIGINAL PAPER
A Review of the Tensile Strength of Rock: Concepts
and Testing
Matthew A. Perras • Mark S. Diederichs
Received: 4 November 2013 / Accepted: 22 January 2014 / Published online: 30 January 2014
 Springer International Publishing Switzerland 2014
Abstract A review of the tensile strength of rock
was conducted to determine the relationship between
direct tensile strength (DTS) and Brazilian tensile
strength (BTS) and to examine the validity of
estimating tensile strength from other measured prop-
erties, such as the crack initiation (CI) threshold. A
data set was gathered from the existing literature
where tensile values could be reliably correlated with
unconfined compressive strength or CI values. It was
determined that the BTS obtained in standard testing is
generally greater than the equivalent DTS and that this
relationship is rock type dependent. CI yields a
reasonable estimate of tensile strength and this corre-
lation is improved when the BTS values are reduced to
DTS values by rock type specific correlations. The
factor f, in DTS = f BTS, can be considered to be
approximately 0.9 for metamorphic, 0.8 for igneous
and 0.7 for sedimentary rocks. The relationships
presented demonstrate that there is wide scatter in
the available data for estimating tensile strength likely
due to both specimen variability and testing configu-
ration, including platen geometry and relative stiff-
ness. Estimates of tensile strength should only be used
for preliminary design purposes and measurements
M. A. Perras (&)  M. S. Diederichs
Queen’s University, Miller Hall, 36 Union Street,
Kingston, ON K7L 3N6, Canada
e-mail: mperras@ethz.ch
M. S. Diederichs
e-mail: mdiederi@geol.queensu.ca
should be made to confirm preliminary assumptions
for each design.
Keywords Laboratory testing  Direct tensile 
Indirect tensile  Brazilian tensile  Unconfined
compressive strength  Crack initiation
1 Introduction
Despite the importance of the tensile capacity in
controlling many failure processes, tensile strength
determination is often overlooked in engineering
practice due to difficulties with obtaining reliable
results. The initiation of fractures in brittle materials
can be a tensile phenomenon as indicated by many
researchers (Haimson and Cornet 2003; Myer et al.
1992; Stacey 1981; Taponnier and Brace 1976;
Griffith 1921). Therefore the tensile strength is an
important aspect of the resistance to failure of a rock or
rock mass. Diederichs and Kaiser (1999) stated that
tensile strength is an important controlling property in
critical span stability of underground openings. The
focus of engineering studies is often on the unconfined
compressive strength (UCS) of intact rock specimens.
Direct tensile strength (DTS) testing is rarely carried
out because of the difficulties in preparing the
specimens; many poorly-prepared specimens fail
invalidly (not through the middle of the specimen)
123
526
and thus must be discarded. Indirect tensile methods,
such as the Brazilian Tensile Test, are much easier to
prepare although invalid tests (fracture is not through
the middle of the specimen or visible platen effects)
are also frequent and each specimen should be
examined after testing to determine its validity
(Colback 1966).
An accurate and representative direct tensile
strength value (DTS) forms the anchor for most rock
strength envelopes defined in stress space. The Hoek
and Brown (1997) formulation, for example, origi-
nates at a tensile strength value (rt = -DTS =
-UCS/mi where mi is a slope parameter for shear
strength). Other methods may use a tensile cutoff limit
in addition to a separate shear strength envelope.
Given the widespread difficulty of determining DTS,
reported in the literature (e.g. Berrenbaum and Brodie
1959; Mellor and Hawkes 1971), the authors reviewed
methods for obtaining tensile strength indirectly. The
most commonly used indirect method is the Brazilian
tensile strength (BTS) test. The review focuses on the
relationship between DTS and BTS for a variety of
rock types found in the literature. Other methods of
estimating tensile strength are also evaluated.
Tensile strength is also closely related to the stress
threshold for fracture initiation in compression. This
limit has been called the crack initiation (CI) limit or
CI (Diederichs and Martin 2010). The estimation of
crack damage thresholds (other than peak strength)
using strain or acoustic emission (AE) based methods
is becoming more common for routine UCS testing as
it can be done with little additional effort during
testing. Diederichs and Martin (2010) summarized the
available methods for measurement of crack damage
thresholds, which includes CI and crack propagation
(CD). These values are important parameters for input
into numerical models. In particular the damage
initiation and spalling limit (DISL) approach of
Diederichs (2007) requires the input of UCS, CI, and
tensile strength to describe the rock mass failure
envelop. The DISL approach is a method of deter-
mining the peak and residual failure envelops to
capture the brittle behaviour of rock masses around
underground excavations. UCS and CI values are
more commonly available or measured; however,
tensile strength is often overlooked. The relationship
between tensile strength, CI and excavation damage
will be explored further.
123
Geotech Geol Eng (2014) 32:525–546
2 Measuring Tensile Strength in the Laboratory
Measuring the tensile strength of rock is governed by the
ISRM (1978) standard which outlines both direct and
indirect Brazilian test methods. ASTM also has standards
governing direct (ASTM 2008a) and indirect Brazilian
(ASTM 2008b) methods. These standards specify, for
example, cylindrical specimens with Height to Diameter
ratios between 2.5 and 3.0 with cemented end caps for
direct tests, and cylindrical discs with a Radius to
Thickness ratio of 1.0 for Brazilian tests. As these are the
standard tensile test methods they are discussed in detail
as well as variations on these methods, which have been
presented in the literature. For completeness other
indirect methods are discussed briefly.
2.1 Direct Tensile Strength
To date direct tensile testing is regarded as the most
valid method for determining the true tensile strength
of rock since there are minimal outside influences
when the test is completed properly (Hoek 1964).
Brace (1964) described the best shape for direct tensile
specimens to be the dog bone shape, as illustrated in
Fig. 1a, where the Height, H, to Diameter, D, ratio
should be 2.0–3.0 of the central test region. The
Radius, R, of curvature of the fillets should be
approximately 1–2 9 D. Specimens of this shape
can be tested with grips, as illustrated in Fig. 1a, which
pull against the lip of the dog bone ends or simply grip
the vertical edge of the specimen, either mechanically
or with adhesive. It should be noted that it is the curved
radius of the dog bone shape which reduces stress
concentrations at the ends of the specimen. In this
review the authors have observed that square dog bone
specimens are also sometimes used. The square shape
can still result in stress concentrations at the ends of
the specimen during loading and invalid failure away
from the central area of the specimen can still occur.
Gripping a cylindrical specimen (non-dog bone)
specimen can cause stress concentrations at the ends
near the grips. A valid direct tensile test should result
in failure at the midpoint of the specimen. With stress
concentrations at the ends of the specimen failure can
initiate near the grips resulting in an invalid test. In
work by Fairhurst (1961) the stress concentrations
were reduced by directly cementing end caps of the
same diameter to the specimen ends (Fig. 1b), which
Geotech Geol Eng (2014) 32:525–546
Fig. 1 Direct tension
testing arrangements using
a split grips for dog-bone
shaped specimens (shape
described by Hoek 1964),
b glued end caps for
cylindrical specimens
(ISRM 1978), c biaxial
extension after Brace (1964)
and, d compression to
tension load converter, for
tensile testing after Gorski
(1993) and Klanphumeesri
(2010). Fa applied force,
R radius, D diameter, A area,
and P water pressure
results in a similar stress distribution as in a uniaxial
compression test. The difficultly with the fixed grips or
end caps is that the specimen can fail in torsion, as the
load transfer mechanism twists during loading. This is
mostly overcome by using universal or ball joints to
allow the specimen to center itself during loading.
These difficulties are avoided in biaxial extension tests
since end grips are eliminated from the experimental
procedure.
2.1.1 Biaxial Extension
Hoek (1964) and Brace (1964) were able to make the
axial stress tensile using dog-bone shaped specimens
in a triaxial cell, where r1 = r2 [ r3. Jaeger and
Cook (1969) state that in practice both r1 and r3 are
increased together to confine the specimen and that the
axial stress is then decreased until failure occurs.
Utilizing the dog bone shape, r3 then exerts a pressure
on the curved portion of the specimen and causes
extension. The test setup is illustrated in Fig. 1c. The
tensile strength can be calculated using Eq. (1);
Fa PðA2  A1 Þ
rt ¼  ð1Þ
A1 A1
where Fa is the applied axial load, A1 is the narrow
diameter, A2 is the head diameter, and P is the
confining pressure. It should be noted that biaxial
extension is not a purely uniaxial tensile test; although
Brace (1964) states that the dog-bone shape specimen
527
yields more reliable data over other tensile methods. If
no axial force (Fa) is applied then the arrangement
becomes a uniaxial tensile test (Hoek 1964). In
addition it is important that the data are plotted on
the appropriate stress path, which in turn is a function
of the curvature of the fillet.
In many laboratories tensile (pull) loading frames
or confining cells are unavailable. Research has been
conducted by others (Gorski 1993; Klanphumeesri
2010) to utilize compression load frames to convert
the downward compression into an upward pull on
dog-bone shaped specimens.
2.1.2 Compression Load Converters
Gorksi (1993) patented a device which allows the
conversion of the compressive load to pull a dog bone
shaped specimen apart in tension. This device can
incorporate the curved dog bone shape recommended
by Brace (1964) and Hoek (1964). A study by
Klanphumeesri (2010) demonstrated the use of a similar
load converter for direct tensile testing. The device used
by Klanphumeesri (2010), illustrated in Fig. 1d, is
simpler to construct than the device of Gorski (1993)
and specimen installation is easier. Even small time
savings during specimen preparation and installation
into the loading frame can increase the number of
specimens which a laboratory can test within practical
time constraints. Lathing specimens to achieve the
desired testing shape remains the limiting factor.
123
528 Geotech Geol Eng (2014) 32:525–546

In all cases of direct tensile testing the tensile accounted for a load of finite width on a solid disk,
strength can be directly measured as the load at with the stress components given by Eqs. (3) and (4),
failure, F, divided by the cross sectional area, A, of modified from Mellor and Hawkes (1971), such that
the failure zone, with the exception of biaxial tension is negative
extension. This is the ideal method of determining
8 h  r 2 i
the tensile strength of rock, however, preparation of < 1  sin2a
the specimens can be time consuming and difficult, Fa R
r1 ¼   r 2  r 4
pRta :1  2
especially with weak rocks. In some cases, such as R cos2a þ R
" #) ð3Þ
shale, the specimen will be unable to withstand  r 2
shaping on a lathe to create the recommended dog 1 1 þ R
tan  2 tana
bone shape. Due to the large variability of tensile 1  Rr
strength of rock specimens, even of the same rock 8 h  2 i
type, many tests are recommended (Jaeger and Cook Fa < 1  Rr sin2a
r3 ¼ þ    
1969). To meet this recommendation it is most pRta :1  2 r 2 cos2a þ r 4
R R
practical, in terms of preparation time, to use grips or " #) ð4Þ
 r 2
glued end caps on straight specimens or to switch to 1þ R
þtan1  2 tana
an indirect testing method. 1  Rr

2.2 Brazilian Tensile Strength
The BTS is determined by an indirect testing method
governed by the ISRM (1978) and ASTM (2008b)
standards, which state that the stress at failure, rt, is a
function of the applied load Fa, the diameter D and the
thickness t at the center of the specimen and can be
determined by Eq. (2).
2Fa Fa
rt ¼ ¼ 0:636 ð2Þ
pDt Dt
Li and Wong (2012) documented the development of
this method starting with the independent proposals
for concrete by Carneiro (1943) and Akazawa (1943).
The first application for rock was by Berrenbaum and
Brodie (1959). Recent studies focus on applications
and improvements through both laboratory and
numerical studies (Tavallali and Vervoort 2010;
Markides et al. 2012; Erarslan and Williams 2012;
Li and Wong 2012). Many studies (see Li and Wong
2012) have focused on the loading platens and contact
area for BTS testing to determine the optimum load
transfer to the specimen such that tensile stresses
develop evenly in the central region and minimize
crushing at the edge of the specimen. The ISRM
(1978) standard suggests a curved set of jaws with a
radius of 1.5 9 the specimen radius; however, there
are a number of other different loading platen
arrangements in use for the BTS test and some of the
most common are shown in Fig. 2. Hondros (1959)
where Fa is the applied load, R is the disc radius, t is
the disc thickness, 2a is the angular width where the
load is applied over (radially), and r is the distance
from the center of the disc. r1 is the vertical stress
within the specimen and r3 is the horizontal stress
within the specimen during loading. The sensitivity of
the normalized stress to the angular width is demon-
strated in Fig. 3a, where an increasing angular width
decreases the normalized tensile stress at the center of
the disk. A closer examination of the tensile stresses
(Fig. 3b) shows that there is approximately an error of
10 % between a point load and a width of 2a = 30.
Also, as the angular width increases the distance (r)
over which the tensile stress is constant decreases.
Theoretically a point load will give the lowest tensile
strength; however, in practice a point load will cause
crushing at the point of contact with the specimen.
Examining the stress difference (Fig. 3c) along the
loading axis shows that it is greatest just inside the
specimen near the platen and the magnitude of the
difference increases as the angular width decreases.
Further from the edge of the specimen, in the central
region, the stress difference becomes similar, inde-
pendent of the loading width. Griffith’s (1921, 1924)
theory and the expanded three dimensional version
suggested by Murrel (1963) can be used to explain the
limiting loading angle width (2a) to minimize CI of
the specimen in the platen area. The equations relating
the principal stresses (r1, r3) and tensile strength (rt)
r in Eqs. (5) and (6) are as follows;

123
Geotech Geol Eng (2014) 32:525–546
Fig. 2 Typical loading
platen arrangements for
BTS testing using a flat
loading platens, b flat
platens with cushion (often
wood), c flat loading platens
with small diameter rods, or
d curved loading jaws
(ISRM 1978). Fa applied
force and D diameter
 
8rt 1 þ rr31
r1 ¼  2 ð5Þ
1  rr31
 
12rt 1 þ 2 rr31
r1 ¼  2 ð6Þ
1  rr31
after Griffith (1921, 1924) and Murrell (1963), respec-
tively. Diederichs (1999) reiterated that Griffith’s
theory is a damage initiation threshold over the full
range of confining stress and that while in compression,
crack accumulation or propagation is required (after
initiation) to fully fail a crystalline rock sample. In
tension the damage initiation and peak strength are
coincident due to unstable CD. In an unconfined test
(r3 = 0) these equations reduce to r1 = -8rt and
r1 = -12rt, respectively. In a DTS test CI and peak
tensile strength should occur almost simultaneously and
therefore r1 = CI. Following from this CI can occur in
the platen area when 2a \ 16 since the stress difference
will exceed CI = 8 rt or CI = 12 rt, according to
Griffith (1921, 1924) or Murrell (1963), respectively, as
indicated in Fig. 3c.
Even with a small distributed load during BTS
testing, crushing in the platen area can occur for weak
specimens. The authors have observed that laboratories
use a cushion, such as plywood or cardboard, to prevent
contact crushing in the platen area of weak specimens.
It should be noted however a cushion that spreads out
during loading can also prematurely crack the specimen
in tension. For the Brazilian Tensile test on a specimen
to be valid the fracture should start and pass from the
529
central region of the specimen out towards the loading
platens. Invalid tests often occur in metamorphic or
sedimentary specimens; where the fractures deviate
along the fabric plane or when fracturing begins in the
platen area. Mixed failure modes can also occur where
fractures propagate to the platens and other locations.
A variation of the Brazilian test, originally devel-
oped to overcome a tendency for mixed failure modes
in solid disks, is the Ring Test. The tensile strength of a
ring test can be determined by Eq. (7),
2Fa Kf
rt ¼ ð7Þ
pDt
where Fa is the applied load, D is the disk diameter, t is
the disk thickness and Kf is the stress concentration
factor. This factor can be calculated using Eq. (8).
 r 2
Kf ¼ 6 þ 38 ð8Þ
R
where r is the hole radius and R is the disk radius. This
type of test is now infrequently used; however other
tensile tests are still in use or are being more recently
developed to overcome the difficulties with direct and
indirect methods.
2.3 Alternative Tensile Testing Methods
An exhaustive review of other tensile tests is beyond
the scope of this paper. Vutukuri et al. (1974) give a
thorough review of early methods used for tensile
strength determination. Several of the best stud-
ied tests, including those more recently investigated
(Luong 1990) are illustrated in Fig. 4, which shows the
123
530 Geotech Geol Eng (2014) 32:525–546

Fig. 3 a Distribution of
normalized stresses with b a
close up view of the tensile
(-) region and c the stress
difference along the loading
diameter (r0?R) of a solid
disc of unit thickness (t) for
different loading widths
(2a), after Hondros (1959).
R radius of specimen,
r distance from center of
specimen, P force, rvert
vertical stress in the
specimen (r1), rhoriz
horizontal stress in the
specimen (r3), CI crack
initiation threshold, and rt
tensile strength

123
Geotech Geol Eng (2014) 32:525–546
Fig. 4 Alternative indirect tensile testing methods, including
a the sleeve fracturing test, b the beam bending test, and c the
modified tension test, as described by Franklin and Dusseault
(1989) and Luong (1990). P water pressure, r inner radius (r2),
R outer radius (r1), rt tensile strength, Fa applied force, L length,
h height, and b thickness (into page)
relationships between the test parameters and the test
specific tensile strength. The sleeve-fracturing test is a
variation of a dilatometer test which is used in situ to
determine the deformability of a rock mass, where an
internal pressure is applied until the specimen splits
radially (Fig. 4a). There are many different types of
beam bending tests used in practice, most of which are
a variation of a three or four point bending test, as
531
illustrated in Fig. 4b, with the associated governing
equations for tensile strength determination. The most
recent development in tensile strength testing is the
modified tension test developed by Luong (1990) in
which a tensile zone develops during loading of an
over cored specimen in the rock bridge, as illustrated
in Fig. 4c. Despite extensive testing with alternative
tensile testing methods for research purposes, in
practice the tensile strength is in some cases estimated
from other index test values. The rest of this paper will
focus on BTS as it is routinely used in industry for
tensile strength determination. As DTS is considered
to be the true tensile strength of rock this paper will
compare test results from the literature with BTS.
Since there is a large volume of BTS test results in the
literature, these are compared with other routine
testing results or estimates. It will be shown that
reasonable estimates of the true tensile strength of rock
can be determined by reducing BTS values or
estimated using CI.
3 Estimating Tensile Strength
Rock properties are often correlated between field
measurements and laboratory test results, such that in
similar rock masses preliminary engineering design
can be conducted prior to obtaining laboratory test
results. It should be stressed that these relationships are
for preliminary analysis only and should never replace
laboratory testing to determine tensile strength.
Many authors have made correlations between
UCS and the point load index, Is50, as summarized by
Zhang (2005). Zhang (2005) suggests a correlation
between tensile strength and the point load index and
reports the relation in Eq. 9.
rt ¼ 1:5Isð50Þ ð9Þ
Hoek and Brown (1997) established the material
constant, mi, to describe the relationship between the
principal stresses and the peak strength, as shown in
Eq. (10), for intact rock.
 0:5
r3
r1 ¼ r3 þ UCS mi þ1 ð10Þ
UCS
It is possible, but not recommended, to fit only the
triaxial and UCS data, and then estimate the tensile
strength by calculating mi (Hoek and Brown 1997). If
only UCS testing has been completed and reliable
123
532
tensile testing data are unavailable an estimate can be
made using rt = -UCS/mi, for brittle rocks (Diede-
richs 2007) and determining mi based on Hoek and
Brown’s (1997) recommendations according to the
rock type.
As previously discussed, the tensile strength can be
related to the CI threshold. Standard UCS testing is
typically the first stage of testing on any engineering
project and therefore the tensile strength could be
estimated using CI. If r3 is set to zero in Eq. 5 than
r1 = 8 rt. In tension the CI threshold is equivalent to
the peak strength, so CI = 8 rt. Similarly, if Eq. 6 is
used, CI = 12 rt. The range (8–12) in the relationship
between tensile strength and CI can be generalized
into Eq. 11;
CI
rt ¼ ð11Þ
b
where b is 8 according to the original Griffith (1921,
1924) theory and can be as high as 12 according to the
modified formulation of Murrell (1963) and Jaeger and
Cook (1969).
These estimates should only be used as preliminary
design estimates for tensile strength and are no
substitute for quality testing on intact rock specimens
for tensile strength. As has already been mentioned,
the difficulties with direct tensile testing often mean
that indirect methods are used to determine the tensile
strength. A search of the literature, summarized in
Table 1, was conducted to examine the relationships
between standard properties measured in the lab and
between direct and indirect (Brazilian) tensile testing
methods.
4 Correlations with Tensile Strength
Coviello et al. (2005) compiled various alternative test
results from the literature which were compared to
DTS. These have been re-plotted in Fig. 5, along with
mean testing results from Coviello et al. (2005). The
results indicate that the BTS and the Hydraulic
Fracture (results processed using elasto-plasticity
shown in Fig. 5) tests give the closest approximation
to the DTS. Given the large volume of BTS strength
data in the literature, as evident in Fig. 5 and in
Table 1, the focus of this paper is largely on the
relationship between DTS and BTS. Estimation of
DTS using CI and the point load index is also
123
Geotech Geol Eng (2014) 32:525–546
examined as a preliminary method when tensile test
results are unavailable.
Despite the large collection of laboratory testing
results in the literature it is very difficult to determine
whether different data sets use the same rock block or
core run interval. In order to correlate compressive and
tensile laboratory results it is important that the
specimens came from very close to each other as
natural variation should be minimized. The specimens
which were compared by the original authors were
taken to have originated from close special proximity
to each other. Where comparison was not made by the
original authors, but spatial information was available,
samples from within 10 cm were compared. Still there
is wide scatter in the relationships presented here.
4.1 Crack Initiation under Compressive Loading
Crack initiation (CI) represents the first onset of new
distributed grain scale cracks within the specimen
during testing. Tapponier and Brace (1976) found that
natural micro-cracks below the CI threshold are
generally limited to the grain scale for crystalline
rocks. As noted previously, the CI threshold (CI) under
compression is a robust material property and is
mechanistically related to the DTS (Diederichs 2003).
In contrast, the threshold for critical crack accumula-
tion, propagation and interaction, CD is more associ-
ated with loading conditions and confining stress.
UCS, in turn, is thereafter related to testing geometry,
loading rate and other influences in addition to the rock
properties.
The different methods for determining the damage
thresholds are illustrated in Fig. 6 and are only briefly
discussed here for clarity. The CI threshold can be
determined as the point where the stress–strain
(volumetric or lateral) deviates from linearity (Brace
et al. 1966; Bieniawski 1967; Lajtai and Lajtai 1974).
If there is pre-existing damage (micro-cracks) in the
specimen then the linear elastic behaviour may not be
present or the true on set of CI may not be accurately
defined. Crouch (1970) suggested that the lateral
strain-axial strain plot could be used to determine CI.
However, around CI, during a compression test, both
lateral and axial strains can change in a similar fashion
and therefore mask the true CI threshold because of the
delay in the deviation from linearity. The CI threshold
can alternatively be determined as the deviation from
linearity of the inverse tangent lateral stiffness (ITLS)
Geotech Geol Eng (2014) 32:525–546 533

Table 1 Summary of average filtered rock properties found in the literature with the number of specimens indicated in brackets
Rock type DTS (MPa) BTS (MPa) UCS (MPa) CI (MPa) CD (MPa) References

Amphibolite – 8.9 110.0 64.0 – Cai (2010)

(2) (1) (1)
Aplite – – 309.9 151.4 262.9 Jacobsson (2005)
(5) (5) (5)
Basalt – 5.1 63.1 – – Graue et al. (2011)
(1) (1)
Calcarenite 0.7 0.6 – – – Coviello et al. (2005)
(1) (1)
Chert 34.8 – 610.9 – – Courtesy of Dr. Evert Hoek
(1) (1)
Diorite – 15.7 251.7 117.5 189.2 Jacobsson (2006)
(10) (10) (10) (10)
Dolomite 5.7 8.0 90.4 35.6 75.3 Ramana and Sarma (1987), Gorski et al.
(2) (6) (4) (4) (4) (2009, 2010, 2011), unpublished
Evaporite – – 109.2 47.1 82.9 Gorski et al. (2009, 2010, 2011),
(12) (12) (12) unpublished
Gneiss 8.2 9.8 114.8 53.7 95.5 Hakala and Heikkila (1997, 1998),
(42) (52) (60) (58) (59) Eloranta (2006)
Granite–igneous 6.3 10.3 170.3 70.4 125.4 Alehossein and Boland (2004), Eloranta
(64) (88) (84) (30) (30) and Hakala (1999), Betournay (1983),
Mellor and Hawkes (1971), Diederichs
(1999), couresy of Dr. Evert Hoek,
Mishra and Basu (2012), Ramana and
Sarma (1987), Hardy and Jayaraman
(1970), Gorski and Yu (1991), Heikkila
and Hakala (1998)
Meta-granite – 13.8 244.8 106.4 201.1 Jacobsson (2004, 2005, 2006, 2007)
(167) (72) (75) (75)
Granodiorite – 7.9 132.9 68 104.7 Eloranta and Hakala (1998)
(23) (10) (10) (10)
Meta-granodiorite – 16.0 148.3 63.6 112.4 Jacobsson (2004)
(12) (5) (5) (5)
Gypsum 1.6 1.6 – – – Andreeve (1991a, b)
(2) (2)
Latite – 9.7 126.4 – – Graue et al. (2011)
(1) (1)
Limestone 7.1 6.0 76.9 – – Mellor and Hawkes (1971),
(8) (60) (55) Klanphumeesri (2010), Gorski et al.
(2009, 2010, 2011), Hardy and
Jayaraman (1970), Perras et al. (2013b),
Golder (2011), Lama and Vutukuri
(1978), Bell (1981)
Marble 7.5 10.1 101.6 – – Jaeger (1967), courtesy of Dr. Evert Hoek
(9) (27) and Efimov (2009), Klanphumeesri
(2010)
Massive Sulphide – 11.9 124 84 – Cai (2010)
(2) (1) (1)

123
534 Geotech Geol Eng (2014) 32:525–546

Table 1 continued
Rock type DTS (MPa) BTS (MPa) UCS (MPa) CI (MPa) CD (MPa) References

Meta-monzodiorite – 14.7 185.5 85.5 144.0 Jacobsson (2006, 2007)

(54) (13) (13) (13)
Mudstone – 5.4 41.0 – – Perras et al. (2013a)
(19) (24)
Meta-pegmatite – 11.5 217.4 103.3 166.5 Jacobsson (2005, 2006), Cai (2010)
(19) (14) (14) (12)
Peridotite – 7.3 111.0 61.0 – Cai (2010)
(4) (2) (2)
Quartzite 16.3 13.0 172.0 78.0 – Courtesy of Evert Hoek Cai (2010),
(2) (3) (3) (2) Ramana and Sarma (1987)
Sandstone 5.1 9.5 105.3 – – Jaeger (1967), Courtesy of Dr. Evert
(11) (47) (50) Hoek, Mishra and Basu (2012),
Klanphumeesri (2010), Hardy and
Jayaraman (1970), Graue et al. (2011),
Perras (2009), Pandey and Singh (1986)
Schist 13.3 11.8 51.7 52.3 – Ramana and Sarma (1987), Mishra and
(7) (33) (26) (6) Basu (2012), Cai (2010)
Shale 5.6 5.9 44.2 18.8 34.5 Hansen and Vogt (1987), Grasle and
(3) (20) (102) (78) (61) Plischke (2010), Gorski et al. (2009,
2010, 2011), Lo and Hori (1979), Perras
(2009), unpublished
Siltstone – 8.2 53.1 17.3 31.2 Perras (2009), Gorski et al. (2009, 2010,
(18) (42) (23) (14) 2011)
Slate – 21.7 216.2 – – Dan et al. (2013)
(1) (1)
Sulphide mix – 5.5 89.8 42.9 – Cai (2010)
(8) (4) (4)
Syenite 20.5 21.1 – – – Andreeve (1991a, b)
(1) (1)
Meta-tonalite – 15.3 169.6 79.0 139.3 Jacobsson (2004)
(14) (4) (4) (4)
Trachyte 13.7 7.7 70.5 – – Jaeger (1967), Graue et al. (2011)
(1) (3) (2)
Values are for results which can be reliably correlated between tensile strength (DTS and BTS) and the other strength thresholds
(UCS, CI and CD). Note that the individual values are used in the figures and not the average values

versus applied stress plot, after Ghazvinian et al.
(2011). The ITLS method avoids the problems stated
above by examining the rate of change in the slope
itself. The CD threshold is the point where cracks
begin to occur in an unstable manner (Bieniawski
1967) and the cracks begin to interact. CD can be
determined, in an unconfined compression test, as the
deviation from the linear elastic response of the strain
in the direction of loading. CI and CD thresholds can
also be detected using AE sensors. Changes in the rate
of AEs during testing have been found to correlate
well with the damage thresholds discussed above (e.g.
Scholz 1968; Lockner 1993; Falls 1993; Eberhardt
et al. 1998), as shown in Fig. 6.
The authors conducted a series of numerical anal-
yses to determine the sensitivity of the DISL approach
to the selected tensile strength used as an input. The
mean values of a set of UCS, CI and BTS test results for
a limestone were used and a variety of maximum
tangential stress (rmax) levels. Tensile strengths were

123
Geotech Geol Eng (2014) 32:525–546
Fig. 5 Comparison between DTS and various alternative
indirect tensile testing methods, where 3 PB is 3 point bending,
4 PB is 4 point bending. Alternative tensile test data compiled
from Coviello et al. (2005), with data from Hardy and
Jayaraman (1970), Jaeger (1967), Jaeger and Hoskins (1966),
and mean values from Pandey and Singh (1986) and
Klanphumeesri (2010). For complete references to data for
BTS and DTS only, see Table 1
also estimated by reduction of the mean BTS value and
using CI/8 and CI/12 (to be discussed later in the
paper). The tensile strength decreases in the following
order; mean BTS, reduced BTS, CI/8, and CI/12. The
stress evolution at the notch tip is such that the rock
mass failure envelope is cross with a positive r3. With
all other parameters being equal, decreasing the tensile
strength causes an increase in the slope of the failure
envelope for positive r3 values.
The study examined the change in the excavation
damage zones (EDZs) with different tensile strengths.
The EDZs were determined for each model, following
the methodology of Perras et al. (2012). The outer
EDZ (EDZo) was determined as the maximum limit of
the plastic yield zone. The inner EDZ (EDZi) was
determined as the volumetric extension-compression
transition. The Highly Damaged Zone was determined
as a low confinement area with rapidly increasing
volumetric and maximum shear strain toward the
excavation boundary. The results of the study are
shown in Fig. 7, which indicates that there can be up to
a difference of 20 % of the predicted damage zone
dimension. The increased slope with lower tensile
535
strength (e.g. CI/12) values actually causes the dam-
age zone dimensions to be smaller than when using a
higher tensile strength (e.g. mean BTS). This could be
a difference of between 0.5 and 1.0 m for a typical
access shaft.
For nuclear waste disposal underground, the dam-
age zone dimension needs to be predicted with a high
degree of confidence so that mitigation measures can
adequately be designed. The typical mitigation mea-
sure envisioned is to create a slot perpendicular to the
excavation axis which cuts off flow along the EDZ
(flow parallel to the excavation axis). If the slot is too
shallow, the cut off slot will be ineffective and if it is
too deep there is a risk of causing more damage to the
rock mass, which may create a short circuit at the cut
off slot tip. This numerical study raises the question
what is the most appropriate tensile strength value to
include in an analysis and whether there are reliable
methods of estimating tensile strength when data is
unavailable.
4.2 Relationships between Intact Rock Properties
In an effort to determine the best relationship between
DTS and other more commonly determined laboratory
parameters the authors started with UCS. The com-
parison between UCS and BTS or DTS data shows
(Fig. 8) wide scatter, as previously mentioned. This is
particularly true for the indirect Brazilian method
(BTS). There is a better correlation with DTS although
there is a similar range of variability and fewer data in
the literature. The mean mi range for different rock
types, 4–32 (Hoek and Brown 1997), should bracket
the data. Examining Fig. 8 the ratio mi = UCS/rt does
bracket the minimum and maximum limits of the data,
approximately. When the information is included in
the source publications, and the data can be separated
by rock types, there is good agreement with the ratio
UCS/rt for different mi values (Fig. 9). If the data are
not separated by study and rock type, then there is less
alignment with different mi values, which can be seen
by the variability in the granite points in Fig. 9. This is
because of the natural variation in granites or other
rock types from different locations, such as grain size
or mineralogical variations, for example.
CI and CD can also be determined from a UCS test
by using strain measurements and/or AE sensors
following the methods described previously. CI is an
important parameter in the DISL method established
123
536 Geotech Geol Eng (2014) 32:525–546

Fig. 6 Different methods of determining crack initiation (CI) and crack propagation (CD) thresholds during a compression test. CC
crack closure, Vol volumetric, ITLS inverse tangent lateral stiffness, and AE acoustic emission

by Diederichs (2007) and critical for understanding
the behaviour of brittle rocks. There is a well-
established correlation between CI and UCS demon-
strated by the work of Brace et al. (1966), which
ranges between 0.3 \ CI/UCS \ 0.5. The data
collected as part of this study is within this range
(between a ratio of 0.4 for CI and 0.8 for CD as seen in
Fig. 10a). Segregating the CI values by main rock
types shows that there is close agreement between the
CI ratio for sedimentary and igneous rocks, with ratio
of 0.42 and 0.43, respectively (Fig. 10b). The meta-
morphic results indicate a slightly higher ratio of 0.45.
CI is more closely related to tensile strength than CD

Fig. 7 The influence of different tensile strengths on the

dimensions of the outer excavation damage zone (EDZo), the
inner excavation damage zone (EDZi) and the highly damage
zone (HDZ), where BTS is the mean Brazilian Tensile Strength
and fBTS is the reduce Brazilian Tensile Strength determined
from the current study Fig. 8 The relationship between UCS–BTS and UCS–DTS

123
Geotech Geol Eng (2014) 32:525–546 537

Fig. 9 The relationship between UCS–BTS and UCS–DTS

sorted by rock type and study

since it is the point in a UCS or triaxial test where

cracks begin to develop in the specimen.
It should be noted that this data set is not a complete
collection of UCS, CI, and CD measurements from the Fig. 10 The relationship between UCS and CI for a the main
literature and is only based on measurements for data set and b the main rock types; metamorphic, sedimentary
which corresponding tensile strength was available. and igneous. The subscript meta is for metamorphic, sed for
sedimentary, and ig for igneous
Based on the available data for the igneous specimens,
a linear correlation which passes through the origin
does not fit the data. The trend is left to illustrate that
the homogeneity of the igneous specimens influences
the behaviour and generates a more consistent CI
value than metamorphic or sedimentary rocks.
There are very few cases for which the spatial
criteria were met between UCS or CI and DTS (only
six tests for CI vs DTS, not shown). The importance
and use of CI has only more recently been established
and as such its routine measurement is still absent in
many laboratories. There are 104 points, mostly from
nuclear waste investigation programs, where stress–
strain plots are available and CI can be determined
from the published result or in some cases CI was
calculated as part of the testing program. A compar-
ison between BTS and CI indicates a general trend
where BTS = 0.14 CI with a weak correlation Fig. 11 Relationship between CI and BTS with Griffith’s
(R2 = 0.11), as shown in Fig. 11. Despite the weak (1921, 1924) 2D (CI/8) and Murrell’s (1963) 3D (CI/12) limits
correlation coefficient it is interesting to observe that based on the main rock types

123
538
Fig. 12 The relationship between CI and BTS for specific
rocks. Note that meta-rocks were identified as metamorphic
where available in the literature or as evident from photographs
of the tested specimens
the relationship closely corresponds to the tensile
strength estimated by Eq. (11) for a dense lower bound
(Griffith’s (1921, 1924) rt = CI/8). A large portion of
the data set is for metamorphic rocks, as shown in
Fig. 11, which shows a large variation. There is
insufficient data for sedimentary and igneous rocks to
establish specific trends. Figure 12 indicates that the
sedimentary rocks are all dolomite and that the
igneous rocks are either granite or granodiorite.
The test results for the sedimentary rocks, in Fig. 12,
are unpublished and came from within the same
borehole, although from different formations suggest-
ing that there is a steep slope to the dolomite CI versus
BTS relationship. For the igneous rocks, the narrow
window of CI values results from the fact that the data
all comes from Finland, for CI versus BTS, and is either
granodiorite or granite, as shown in Fig. 12 (Heikkilla
and Hakala 1998; Eloranta and Hakala 1999). The
similarity in grain type and size may result in similar CI
values, similar to the UCS relationship, suggesting that
for homogenous rocks, CI is a material property. This
concept is further reinforced on closer inspection of the
individual rocks in Fig. 12, which show alignment with
common CI values for potentially more homogenous or
uniform specimens at the testing scale, such as schist
and massive sulphide. Other rocks which are typically
considered heterogeneous by nature show more scatter,
such as sulphide mix, pegmatite, and meta-granite.
Further testing is needed to confirm this finding.
Tensile strength determined from direct and indi-
rect methods are seldom equivalent. Direct tensile
123
Geotech Geol Eng (2014) 32:525–546
Fig. 13 The relationship between BTS and DTS for a the
unfiltered and filtered data and b for the filtered data with the
main rock types. Note that the data were filtered based on the
specimen geometry. The subscript Filtered means the best fit
line is through the filtered data only (diamonds) and All means
the best fit line is through all the data (squares). The subscript
meta is for metamorphic, sed for sedimentary, and ig for igneous
testing is considered to yield the true tensile strength of
intact rock whereas the Brazilian tensile test is
considered to overestimate the true tensile strength
of intact rock. A general rule of thumb used in practice
is that DTS = 0.67 BTS or two thirds. If the complete
data set for DTS and BTS values is considered, the
linear regression yields the general rule of thumb
closely (DTS = 0.65 BTS) with a coefficient of
determination of 0.42, as shown in Fig. 13a. The
small red squares in Fig. 13a are a series of tests
reported by Gorski et al. (2007) on meta-granitoid
specimens from the Forsmark Nuclear Waste storage
investigations in Sweden. These specimens were
prepared without the recommended fillets (i.e. they
were square dog bone in shape) at the ends of the
specimens (Hoek 1964; Brace 1964). If these points
Geotech Geol Eng (2014) 32:525–546
Fig. 14 The relationship between BTS and DTS for a specific
rocks and b for the degree of foliation. The subscript F foliation,
M moderate foliation, W weak foliation, and NF no foliation
are filtered from the data set a better coefficient of
determination, 0.71, is achieved from a linear fit,
where DTS = 0.85 BTS. The only better fit is
achieved with a second order polynomial with a
correlation of 0.72. Since there is only a marginal
increase in the correlation coefficient when using the
polynomial, the linear fit will be used for the
remainder of the analysis.
By examining the relationship between BTS and
DTS by main rock types (see Fig. 13b), slightly better
fits can be achieved for igneous and sedimentary
rocks. However, the correlation with metamorphic
rocks is slightly lower than the filtered data set as a
whole. The fits based on the main rock types indicate
that there is a larger difference between DTS and BTS
values for sedimentary rocks (DTS = 0.68 BTS)
followed by igneous with slightly less difference
(DTS = 0.86 BTS) and metamorphic rocks are close
to a one to one relationship, with DTS = 0.93 BTS.
539
Fig. 15 CI compared to DTS calculated from the BTS test
results using the linear relation of DTS = f BTS where
a generally f = 0.84 and b specifically f = 0.93 for metamor-
phic, 0.86 for igneous, and 0.68 for sedimentary rocks
The influence of sedimentary bedding or metamor-
phic foliation can cause larger variation in the tensile
strength and the orientation of the fabric during testing
will also influence the test result. Dan et al. (2013)
observed that the largest influence occurs when the
fabric is oriented between 30 and 60 from the loading
direction for Brazilian tests. Orientation of fabric for
the collected data is seldom indicated, however; a
degree of foliation can be implied by the specific rock
type. The DTS and BTS data, in Fig. 14a, was first
broken down into specific rock types, which shows that
for a highly foliated rock such as schist that there is a
uniform one to one relationship between BTS and
DTS. Other rock types which could be considered to be
less foliated than a schist, such as sandstone and
marble, show less steep trends. The granite points are
variable and speculatively dependent on grain size and
variations in mineralogy, although information is
lacking in the literature in this regard. To examine this
123
540
further, the data were ranked according to the degree of
foliation as follows;
• No Foliation means there is no distinct orientation
of minerals or separation implied or indicated.
• Weak means that there is a weak alignment of
minerals without continuous bands.
• Moderate means that there is an alignment of
mineral and separation of like minerals forming
discontinuous bands.
• Foliated means that there is an alignment of
minerals and separation of like minerals forming
continuous and distinct bands.
Figure 14b shows that foliated specimens have the
closest relationship between BTS and DTS, where
DTS = 0.96 BTS, followed by moderate and non-
foliated with DTS = 0.83 BTS and that weakly
foliated specimens have the lowest relationship with
DTS = 0.70 BTS. These relationships are very similar
to the main rock type relations, which appear to
capture the degree of foliation adequately, with
metamorphic rocks being closest to a one to one
relationship between BTS and DTS.
As previously discussed the relationship between
CI and BTS (Fig. 11) has a weak coefficient of
determination of 0.11. If the BTS values are reduced
using DTS = 0.85 BTS, the linear relation for all the
data, there is a slight change in the relationship, where
DTS = 0.12CI (Fig. 15a). There is no change in the
coefficient of determination since the values are
simply reduced linearly by the same amount in all
cases. However; if the main rock type relationships
between BTS and DTS are used from Fig. 13 then
there is a slight increase in the coefficient of determi-
nation to 0.22, as shown in Fig. 15b. The relationship
than becomes DTS = 0.13 CI, which corresponds to
Eq. 10 with b = 8.
Field tests are a desirable way to quickly assess
the suitability of a rock mass for construction
purposes. The point load index can give a quick
estimate for preliminary design purposes. A com-
parison between the tensile strength estimated by
the point load method (using Eq. 9) and BTS results
are shown in Fig. 16. The normal fits to both
methods agree very well, with the mean Point Load
and BTS values being 6.3 and 6.0 MPa, respec-
tively. The coefficient of variation (COV) is high,
but within the acceptable levels for geomaterials,
according to Langford (2013). There are many
123
Geotech Geol Eng (2014) 32:525–546
Fig. 16 Comparison of histograms for BTS and tensile strength
estimated from point load measurement of the Quintner
Limestone from Sargans, Switzerland
influences which can contribute to the variability of
the test results.
4.3 Variability in Test Results
Three main sources of variability within the data
presented are spatial relationships between specimens
compared, testing quality control and procedures, and
natural heterogeneity.
The authors used two criteria to compare the results
of different test methods to ensure that the results are
representative of the same rock type. Firstly the
authors looked at testing results which had direct
comparisons published by the original researchers.
Secondly, where information about the spatial loca-
tion, such as borehole number and depth, of the
specimens was provided by the original researchers
specimens that were within 10 cm of each other were
considered to be representative of the same rock. This
distance, in drill core for example, is in the approx-
imate range that one would be able to find both a UCS
and a DTS or BTS specimen that are similar in nature.
In most cases spatial correlations in this manner were
confirmed by exmination of photographic evidence
provided in the published testing reports, such as in
Jacobsson’s (2004 through 2007) work.
A large volume of data came from testing for
nuclear waste storage underground and testing results
with different methods were often published in
different reports. In all the collected data, the results
of different test methods were only compared when
Geotech Geol Eng (2014) 32:525–546
Fig. 17 The relationship between a BTS and DTS for granites
from different locations and b CI and DTS = fBTS for meta-
granites from Sweden and Finland
the same author(s) published the compressive and
tensile results or if it could be confirmed that the same
laboratory was used for the different test methods (as
well as meeting the spatial criteria outlined at the start
of Sect. 4). This should ensure consistent care in
specimen preparation and testing procedures. How-
ever, the BTS loading platen could influence the final
tensile strength. Mellor and Hawkes (1971) studied the
influence and determined that a curved loading jig and
a cushion gave reliable and consistent results.
To understand natural variability in the results, a sub-
section of the collected data, granites, is used to discuss
possible natural influencing factors. Natural variability,
such as grain size variation, within a sub-set based on a
specific rock type can account for some of the scatter in
the data presented in this paper. Due to limited amounts
of specific rock type data, it is difficult to determine
relationships between the test results. It is however
possible to demonstrate that there is an influence based
541
Fig. 18 Histogram of the ratio between DTS and BTS for the
main rock types (meta metamorphic, sed sedimentary, ig
igneous, St. Dev standard deviation, and COV coefficient of
variation)
on where the specimens have come from, as shown in
Fig. 17. This graph presents data from a collection of
granite and meta-granite specimens from various loca-
tions. In Fig. 17a there are generally clusters or
alignment of data points based on the location from
which the specimens originated, suggesting that the
relationship between BTS and DTS is influenced by the
nature of the specific granitic properties, such as grain
size or mineralogical percentages. On average, per
location, the result would plot very close to the
DTS = 0.84BTS line, with the exception of the
AECL—420 level specimens.
According to Gorski and Yu (1991) the AECL—
420 level specimens came from vertical boreholes
drilled from within an adit at the Pinawa Underground
Research Laboratory in Manitoba. They concluded
that the specimens must contain micro-fractures from
stress relief during drilling. The micro-fractures would
be oriented perpendicular to the direction of drilling
and this would reduce the DTS, since the specimen
would also be loaded perpendicular to the micro-
fractures. The BTS would be less influenced by the
micro-fractures since the direction of loading would
be parallel to the micro-fractures (see inset diagram in
Fig. 17a). Lim and Martin (2010) analyzed core
specimens from the 420 level and showed that intact
specimens can be partially disked or contain micro-
fractures.
For the relationship between CI and DTS (as a
function of BTS) clustering occurs based on where the
specimens originated from, although less clearly
123
542
(Fig. 17b). The Forsmark samples in Fig. 17b were
divided based on the borehole from which they were
taken which shows some clustering despite all being
metagranitest. The variation of the test results with
depth was examined, but there was not a clear
relationship indicated by the limited data. The clus-
tering suggests that even though the specimens are all
meta-granites, natural variability at the same project
site influences the strength of the material. Further
study is required to explore how natural variability
within a specific rock type influences the relationships
examined in the paper.
Despite the variability of the test results gathered,
when a histogram of the ratio between DTS and BTS is
examined (Fig. 18) there is a clear indication that the
DTS is over estimated by the Brazilian tensile test
method. Although the over estimation may be small for
igneous rocks, which exhibits a strong peak. However,
the peaks for the metamorphic and sedimentary rocks
are less well pronounced, leading to a greater over
estimation. The results for each rock type are normally
distributed. Based on the normal fit the mean values of
the DTS/BTS ratios are 0.86, 0.82 and 0.70 for the
metamorphic, igneous and sedimentary rocks, respec-
tively. The igneous results show the smallest standard
deviation and COV, as expected since there tends to be
more consistent grain size and stiffness in the igneous
rocks tested and reported in the literature. Despite the
variability of the data from the other rock types there is
an indication that the relationship between DTS and
BTS is rock type dependent. Using the above ratios an
approximate estimate of the DTS can be estimated
from BTS test results.
5 Discussion
The standardization of tensile strength testing for rock
mechanics suggests both direct and indirect methods
can be used. In light of the difficulties with direct
tensile methods, indirect methods are more often
conducted. The authors would go a step further and
suggest that indirect specimens be cut from the end of
UCS specimens whenever possible to get the best
possible correlation between tensile and compressive
damage thresholds. The specimens should be exam-
ined carefully in order to ensure that they are of a
similar nature and that the fabric if present is aligned
properly for comparison.
123
Geotech Geol Eng (2014) 32:525–546
Many rocks have a fabric, either metamorphic or
depositional in origin, which influences the strength;
this influence depends on the loading direction with
respect to the orientation of the fabric. The weakest
orientation is typically found between 30 and 60
inclination and a recent study, by Dan et al. (2013),
demonstrates similar results for BTS. The optimum
comparison of tests from different specimens with a
fabric is one in which the failure plane interacts with
the fabric in a similar manner, i.e. a vertical failure
plane cuts across or parallel to the fabric. In the data
collected from the literature, this information was not
always given, and may account for much of the scatter
in both comparing DTS and CI with BTS results.
Observational evidence by other researchers (Grif-
fith 1921, Tapponier and Brace 1976; Stacey 1981;
Myer et al. 1992; Lee and Haimson 1993) suggests that
the initiation of damage is one of extension. In a DTS
test, slip dislocation is not prevalent for most non-
metallic solids (Lockner 1993). If slip dislocation is
not able to occur, in compression of hard brittle rocks,
then shear rupture must also be initiated by tensile
failure at the micro scale. Since CI is a measure of the
stress required to initiate cracking in a compression
test, the theory supports the relationship found in this
paper between the tensile strength and CI. The CI
threshold has been found to be relatively insensitive to
confinement in comparison to the CD threshold
(Diederichs 1999; Martin 1994; Brace et al. 1966;
Pestman and van Munster 1996).
Since the CD threshold is confinement dependent,
crack growth during a Brazilian Tensile test could be
suppressed in the outer 20–25 % of the radial dimen-
sion of the disk when the loading width (2a) is less
than 16 (see Fig. 3a). However; according to the ISRM
(1978) suggested method for the Brazilian test, the
difference between the load at primary fracture and the
ultimate load capacity of the specimen after fracture is
roughly 5 %. This would suggest that suppressed
crack growth is not the leading factor causing BTS to
be larger than DTS, but perhaps it does play a part in
combination with other factors. The physical mecha-
nism of failure is resisted by a small frictional force
between the specimen and the loading platen. There-
fore the specimen is unable to fail in a purely tensile
manner (i.e. similar to a direct test). Markides et al.
(2011) indicate that the frictional resistance only
influences a narrow region near the platen. Another
aspect which could influence the variation in tensile
Geotech Geol Eng (2014) 32:525–546
strength is the mineralogy. The rocks which are
composed of dominantly one type of mineral appear to
plot closest to the DTS = BTS line and those which
are likely to have variation in mineralogy have more
scatter. For those samples which display large varia-
tions in mineralogy it would also be more difficult to
select similar samples even within 10 cm of each
other. More minerals contained within a sample with
different moduli could influence crack propagation by
forcing more tortuous fracture pathways. This would
influence the BTS more, as the fractures must
propagate through zones of confinement due to
differences in grain stiffness and strength. Despite
the influences discussed above there is a clear
indication that the relationship between DTS and
other methods of determining the tensile strength is
rock type dependent.
6 Conclusions
The test results in the literature for tensile strength
suggest that it is difficult to estimate the true tensile
strength accurately from other laboratory results. An
estimate from UCS values gives the most erroneous
results. UCS/rt does give a good approximation for
the Hoek and Brown (1997) constant mi, however,
this relationship should be used with caution to
estimate tensile strength particularly if the mi value
has not been determined from laboratory testing.
Using CI to predict tensile strength generates less
variability than using UCS and is in close agreement
with rt = CI/8. It could be used for preliminary
assessment in engineering projects. Actual testing is
always recommended.
It was shown that the relationship with CI is also
rock type dependent, similar to the relationship
between DTS and BTS. In fact better correlation
coefficients exist when using the main rock types, such
that f in DTS = f BTS ranges from approximately 0.9
for metamorphic, 0.8 for igneous and 0.7 for sedi-
mentary rocks, for practical applications. Reducing
the BTS values based on the main rock types improves
the correlation with CI.
There is a large degree of scatter in the relationships
presented in this paper, which is in part due to the
differences in rock type and mineralogy. Closer
examination of granite and meta-granites indicates
that some of this scatter is also related to natural
543
variability. This variability was examined basd on the
geographic location from which the samples came
from and locally based on different boreholes from the
same site. In both cases the scatter is reduced when the
specific geographic information is used to further sub-
divide the specific rock types (granite or metagranite
examples in Fig. 17).
Further study is required on igneous and sedimen-
tary rocks to compare DTS with BTS and tensile
strength with CI to confirm the findings of this review
of the tensile strength of rock.
Acknowledgments The authors would like to thank the
Nuclear Waste Management Organization (NWMO) of
Canada and the Natural Sciences and Engineering Research
Council (NSERC) of Canada for funding this review. Special
thanks is due to Dr. Evert Hoek for use of testing data and for
discussions related to this paper.
References
Akazawa T (1943) New test method for evaluating internal
stress due to compression of concrete: the splitting tension
test. J Japan Soc Civil Eng 29:777–787
Alehossein H, Boland JN (2004) Strength, toughness, damage
and fatigue of rock. In: Atrens A, Boland JN, Clegg R,
Giffiths JR (eds) Proceedings of the international confer-
ence on structural integrity and fracture. Brisbane, Aus-
tralia, SIF 836. http://espace.library.uq.edu.au/eserv/UQ:
10122/Alehossein_sif04.pdf. Accessed 13 Oct 2013
Andreev GE (1991a) A review of the Brazilian test for rock
tensile strength determination. Part I: calculation formula.
Min Sci Technol 13(3):445–456. doi:10.1016/0167-
9031(91)91006-4
Andreev GE (1991b) A review of the Brazilian test for rock
tensile strength determination. Part II: contact conditions.
Min Sci Technol 13(3):457–465. doi:10.1016/0167-
9031(91)91035-G
ASTM (2008a) D2936-08: standard test method for direct ten-
sile strength of intact rock core specimens. ASTM Inter-
national, West Conshohocken
ASTM (2008b) D3967-08: standard test method for splitting
tensile strength of intact rock core specimens. ASTM
International, West Conshohocken
Bell FG (1981) A survey of the physical properties of some
carbonate rocks. Bull Int Assoc Eng Geol 24:105–110.
doi:10.1007/BF02595261
Berrenbaum R, Brodie I (1959) Measurement of the tensile
strength of brittle materials. Brit J Appl Phys 10:281–286.
http://iopscience.iop.org/0508-3443/10/6/307. Accessed
14 Dec 2012
Betournay M (1983) Examinatin of URL-1, URL-2, and URL-5
Uniaxial compressive and tensile test data. Canadian Centre
for Mineral and Energy Technology, Mining Research
Laboratories Division Report, ERP/MRL 83-26(TR)
Bieniawski ZT (1967) Mechanism of brittle fracture of rock:
part I—theory of the fracture process. Int J Rock Mech Min
123
544
Sci Geomech Abs 4(4):395–406. doi:10.1016/0148-
9062(67)90030-7
Brace WF (1964) Brittle fracture of rocks. In: Judd WR (ed)
Proceedings of the International Conference on State of
Stress in the Earth’s Crust. Elsevier, New York,
pp 111–174
Brace WF, Paulding B, Scholz C (1966) Dilatancy in the frac-
ture of crystalline rocks. J Geophys Res 71:3939–3953.
doi:10.1029/JZ071i016p03939
Cai M (2010) Practical estimates of tensile strength and the
Hoek-Brown strength parameter mi of brittle rocks. Rock
Mech Rock Eng 43:167–184. doi:10.1007/s00603-009-
0053-1
Carneiro FLLB (1943) A new method to determine the tensile
strength of concrete. In: Proceedings of the 5th Meeting of
the Brazilian Association for Technical Rules
Colback PSB (1966) An analysis of brittle fracture initiation and
propagation in the Brazilian test. 1st ISRM congress. Lis-
bon, Portugal, 1CONGRESS-1966-066
Coviello A, Lagioia R, Nova R (2005) On the measurement of
the tensile strength of soft rocks. Rock Mech Rock Eng
38(4):251–273. doi:10.1007/s00603-005-0054-7
Crouch SL (1970) Experimental determination of volumetric
strains in failed rock. Int J Rock Mech Min Sci
Geomech Abs 7(6):589–603. doi:10.1016/0148-9062(70)
90020-3
Dan DQ, Konietzky H, Herbst M (2013) Brazilian tensile
strength tests on some anisotropic rocks. Int J Rock Mech
Min Sci 58:1–7. doi:10.1016/j.ijrmms.2012.08.010
Diedeirchs MS (1999) Instability of hard rockmasses: the role of
tensile damage and relaxation. PhD Thesis, Department of
Civil Engineering, University of Waterloo, Waterloo,
Canada, pp 566
Diederichs MS (2003) Rock fracture and collapse under low con-
finement conditions. Rock Mech Rock Eng 36(5):339–381.
doi:10.1007/s00603-003-0015-y
Diederichs MS (2007) The 2003 Canadian geotechnical collo-
quium: mechanistic interpretation and practical application
of damage and spalling prediction criteria for deep tun-
nelling. Can Geotech J 44:1082–1116. doi:10.1139/T07-
033
Diederichs MS, Kaiser PK (1999) Tensile strength and abutment
relaxation as failure control mechanics in underground
excavations. Int J Rock Mech Min Sci 36:69–96. doi:10.
1016/S0148-9062(98)00179-X
Diederichs MS, Martin CD (2010) Measurement of spalling
parameters from laboratory testing. In: Zhao J, Labiouse V,
Dudt JP, Mathier JF (eds) Proc of Eurock 2010. Lausanne,
Switzerland
Eberhardt E, Stead D, Stimpson B, Read RS (1998) Identifying
crack initiation and propagation thresholds in brittle rock.
Can Geotech J 35:222–233. doi:10.1139/t97-091
Efimov VP (2009) The rock strength in different tension con-
ditions. J Min Sci 45(6):569–575. doi:10.1007/s10913-
009-0071-0
Eloranta P (2006) Posiva laboratory testing report WR-2006-80.
http://www.posiva.fi/
Eloranta P, Hakala M (1998) Posiva laboratory testing report
WR-98-49. http://www.posiva.fi/
Eloranta P, Hakala M (1999) Posiva laboratory testing report
WR-99-47. http://www.posiva.fi/
123
Geotech Geol Eng (2014) 32:525–546
Erarslan N, Williams DJ (2012) Experimental, numerical and
analytical studies on tensile strength of rocks. Int J Rock
Mech Min Sci 49(1):21–30. doi:10.1016/j.ijrmms.2011.11.
007
Fairhurst C (1961) Laboratory measurements of some physical
properties of rock. In: Proceedings of the fourth sympo-
sium on rock mechanics. Pennsylvania, USA
Falls S (1993) Ultrasonic imaging and acoustic emission studies
of microcrack development in lac du bonnet granite. PhD
Thesis, Queens University, Kingston, Canada
Franklin JA, Dusseault MB (1989) Rock engineering. McGraw-
Hill, New York, p 600
Ghazvinian E, Diederichs M, Archibald J (2011) Challenges
related to standardized detection of crack initiation
thresholds for lower-bound or ultra-long-term strength
prediction of rock. In: Proceedings of the Pan-Am CGS
Geotechnical Conference. Toronto, Canada
Golder (2011) Geotechnical data report: geotechnical and
hydrogeological investigation—Ottawa light rail transit
(OLRT) tunnel (segment 2). Ottawa, Ontario. Report
Number 10-1121-0222
Gorski B (1993) Tensile testing apparatus. United States Patent,
5193396
Gorski B, Yu YS (1991) Tensile strength tests on URL rock
specimens from borehole 401-009-HF1. CANMET Mining
Research Laboratories report MRL 91-080(TR)
Gorski B, Conlon B, Ljunggren B (2007) Forsmark Site inves-
tigation—Determination of the direct and indirect tensile
strength on cores from borehole KFM01D. SKB P-07-
76,Svensk ärnbränslehantering AB
Gorski B, Anderson T, Conlon T (2009) DGR site character-
ization documents, technical reports TR-07-03 and TR-08-
11. www.nwmo.ca
Gorski B, Anderson T, Conlon T (2010) DGR site character-
ization documents, technical reports TR-08-24 and TR-08-
36. www.nwmo.ca
Gorski B, Rodgers D, Conlon B (2011) DGR site character-
ization document, technical report TR-09-07 www.nwmo.
ca
Grasle W, Plischke I (2010) LT experiment: mechanical
behavior of Opalinus clay, final report from Phases 6-14.
Mont Terri Project Technical Report 2009–07
Graue R, Siegesmund S, Middendorf B (2011) Quality assess-
ment of replacement stones for the Cologne Cathedral:
mineralogical and petrophysical requirements. Environ
Earth Sci 63:1799–1822. doi:10.1007/s12665-011-1077-x
Griffith AA (1921) The phenomena of rupture and flow in solids.
Philos Trans R Soc Lond 221A:163–198
Griffith AA (1924) Theory of rupture. In: Proceedings of the
1st international congress on applied mechanics. Delft,
pp 55–63
Haimson BC, Cornet FH (2003) ISRM suggested methods for
rock stress estimation—Part 3: hydraulic fracturing (HF)
and/or hydraulic testing of pre-existing fractures (HTPF).
Int J of Rock Mech Min Sci 40:1011–1020. doi:10.1016/j.
ijrmms.2003.08.002
Hakala M, Heikkila E (1997) Posiva laboratory testing reports
WR-97-04, WR-97-07e. http://www.posiva.fi/
Hansen FD, Vogt TJ (1987) Thermo mechanical properties of
selected shales. Oak Ridge National Laboratory Report
ORNL/Sub/85-97343/2 (RSI-0305)
Geotech Geol Eng (2014) 32:525–546
Hardy HR, Jayaraman NI (1970) An investigation of methods
for the determination of the tensile strength of rock. In:
Proceedings of the 2nd congress international society for
rock mechanics, Belgrade, vol 3, pp 85–92
Heikkila E, Hakala M (1998) Posiva laboratory testing reports
WR-98-06e, WR-98-21e. http://www.posiva.fi/
Hoek E (1964) Fracture of anisotropic rock. J S Afr Inst Min
Metall 64(10):501–518. http://www.rocscience.com
Hoek E, Brown T (1997) Practical estimates of rock mass
strength. J Rock Mech Min Sci 34(8):1165–1186. http://
www.rocscience.com
Hondros G (1959) The evaluation of Poisson’s ratio and the
modulus of materials of a low tensile resistance by the
Brazilian (indirect tensile) test with particular reference to
concrete. Aust J Appl Sci 10(3):243–268
ISRM (1978) Suggested methods for determining tensile strength
of rock materials. Int J Rock Mech Min Sci Geomech Abstr
15(3):99–103. doi:10.1016/0148-9062(78)90003-7
Jacobsson L (2004) Site investigation reports. Swedish nuclear
fuel and waste management Co. Technical Reports P-04-
170, P-04-172, P-04-173, P-04-174, P-04-223, P-04-225,
and P-04-226. http://www.skb.se
Jacobsson L (2005) Site investigation reports. Swedish nuclear
fuel and waste management Co. Technical Reports P-05-
97, P-05-98, P-05-120, P-05-121, P-05-211, and P-05-212.
http://www.skb.se
Jacobsson L (2006) Site investigation reports. Swedish nuclear
fuel and waste management Co. Technical Reports P-06-
37, P-06-38, P-06-73, P-06-74, P-06-270, P-06-271, P-06-
299, and P-06-300. http://www.skb.se
Jacobsson L (2007) Site investigation reports. Swedish nuclear
fuel and waste management Co. Technical Reports P-07-
142, P-07-143, P-07-145, P-07-146, and P-07-207. http://
www.skb.se
Jaeger JC (1967) Failure of rocks under tensile conditions. Int J
Rock Mech Min Sci Geomech Abstr 4(2):219–227. doi:10.
1016/0148-9062(67)90046-0
Jaeger JC, Cook NGW (1969) Fundamentals of rock mechanics.
Methuen and Co Ltd., London, p 513
Jaeger JC, Hoskins ER (1966) Rock failure under the confined
Brazilian test. J Geophys Res 71:2651–2659. doi:10.1029/
JZ071i010p02651
Klanphumeesri S (2010) Direct tension testing of rock speci-
mens. Masters of Engineering Thesis, Suranaree Univer-
sity of Technology
Lajtai EZ, Lajtai VN (1974) The evolution of brittle fracture in
rocks. J Geo Soc 130(1):1–16. doi:10.1144/gsjgs.130.1.
0001
Lama RD, Vutukuri VS (1978) Handbook on mechanical
properties of rocks—testing techniques and results 3(1).
Trans Tech Publications, Clausthal
Langford JC (2013) Application of reliability methods to the
design of underground structures. PhD Thesis, Queen’s
University, Kingston, Ontario, Canada
Lee MY, Haimson BC (1993) Laboratory study of borehole
breakouts in Lac du Bonnet Granite: a case of extensile
failure mechanism. Int J Rock Mech Min Sci Geomech
Abstr 30(7):1039–1045
Li D,Wong LNY (2012) The Brazilian disc test for rock
mechanics applications: review and new insights. Rock
Mech Rock Eng May. doi:10.1007/s00603-012-0257-7
545
Lim SS, Martin CD (2010) Core disking and its relationship with
stress magnitude for Lac du Bonnet granite. Int J Rock
Mech Min Sci 47: 254-264. doi:10.1016/j.ijrmms.2009.11.
007
Lo KY, Hori M (1979) Deformation and strength properties of
some rocks in Southern Ontario. Can Geotech J
16:108–120. doi:10.1139/t79-010
Lockner D (1993) The role of acoustic emission in the study of
rock. Int J Rock Mech Min Sci Geomech Abstr
30(7):883–899. doi:10.1016/0148-9062(93)90041-B
Luong MP (1990) Tensile and shear strengths of concrete and
rock. Eng Fract Mech 35(1,2,3):127–135. doi:10.1016/
0013-7944(90)90190-R
Markides CF, Pazis DN, Kourkoulis SK (2011) Influence of
friction on the stress field of the Brazilian tensile test. Rock
Mech Rock Eng 44:113–119. doi:10.1007/s00603-010-
0115-4
Markides CF, Pazis DN, Kourkoulis SK (2012) The Brazilian
disc under non-uniform distribution of radial pressure and
friction. Int J Rock Mech Min Sci 50(1):47–55. doi:10.
1016/j.ijrmms.2011.12.012
Martin CD (1994) The strength of massive Lac du Bonnet
Granite around underground openings. PhD Thesis, Uni-
versity of Manitoba
Mellor M, Hawkes I (1971) Measurement of tensile strength by
diametral compression of discs and annuli. Eng Geol
5(3):173–225. doi:10.1016/0013-7952(71)90001-9
Mishra DA, Basu A (2012) Use of the block punch test to
predict the compressive and tensile strengths of rocks.
Int J Rock Mech Min Sci 51:119–127. doi:10.1016/j.
ijrmms.2012.01.016
Murrell SAF (1963) A criterion for brittle fracture of rocks and
concrete under triaxial stress and the effect of pore pressure
on the criterion. In: Fairhurst (ed) Rock mechanics. Pro-
ceedings of the fifth rock mechanics symposium, Univer-
sity of Minnesota, Oxford, Pergamon, pp 563–577
Myer LR, Kemeny JM, Zheng Z, Suarex R, Ewy RT, Cook
NGW (1992) Extensile cracking in porous rock under
differential compressive stress. In: Li LY (ed) Microme-
chanical modeling of quasi-brittle materials behaviour.
Applied Mechanics Reviews, 45(8):263–280. doi:10.1115/
1.3119757
Pandey P, Singh DP (1986) Deformation of a rock in different
tensile tests. Eng Geol 22(3):281–292. doi:10.1016/0013-
7952(86)90029-3
Perras MA (2009) Tunnelling in horizontally laminated ground:
the influence of lamination thickness on anisotropic
behaviour and practical observations from the Niagara
Tunnel Project. Masters Thesis, Queen’s University,
Kingston, Ontario, Canada
Perras MA, Langford C, Ghazvinian E, Diederichs MS (2012)
Numerical delineation of the excavation damage zones:
from rock properties to statistical distribution of the
dimensions. In: Proceedings of the Eurock, Stockholm,
Sweden
Perras MA, Besaw D, Diederich MS (2013) Geological and
geotechnical observations from the Niagara Tunnel Pro-
ject. Submitted to the Bulletin of Engineering Geology and
the Environment, BOEG-D-12-00133R1
Perras MA, Ghazvinian E, Amann F, Wannenmacher H,
Diederichs MS, (2013b) Back analysis of rock mass
123
546
behavior of the Quintner Limestone at the Gonzen mine
near Sargans, Switzerland. In: The proceedings of Eurock
2013, Wroclaw, Poland
Pestman BJ, Van Munster JG (1996) An acoustic emission study
of damage development and stress-memory effects in
sandstone. Int J Rock Mech Min Sci 33(6):585–593.
doi:10.1016/0148-9062(96)00011-3
Ramana YV, Sarma LP (1987) Split-collar tensile test grips for
short rock cores. Eng Geol 23:255–261. doi:10.1016/0013-
7952(87)90092-5
Scholz CH (1968) Microfracturing and the inelastic deformation of
rock in compression. J Geophy Research 73(4):1417–1432.
doi:10.1029/JB073i004p01417
Stacey TR (1981) A simple extension strain criterion for fracture
of brittle rock. Int J Frac 18:469–474. doi:10.1016/0148-
9062(81)90511-8
Taponnier P, Brace WF (1976) Development of stress-induced
microcracks in Westerly granite. Int J Rock Mech Min Sci
123
View publication stats
Geotech Geol Eng (2014) 32:525–546
and Geomech Abstr 13(4):103–112. doi:10.1016/0148-
9062(76)91937-9
Tavallali A, Vervoort A (2010) Failure of layered sandstone
under Brazilian test conditions: effect of micro-scale
parameters on macro-scale behaviour. Rock Mech Rock
Eng 43(5):641–653. doi:10.1007/s00603-010-0084-7
Vutukuri VS, Lama RD, Saluja SS (1974) Handbook on
mechanical properties of rocks. Series on rock and soil
mechanics 2(1), Trans Tech Publications, Ohio, USA
Zhang L (2005) Engineering properties of rocks (pp: 175-202).
Elsevier. Online version available at: http://www.
knovel.com/web/portal/browse/display?_EXT_KNOVEL_
DISPLAY_bookid=1848&VerticalID=0. Accessed 15 Dec
2012